Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Introduction
A common offshore structure in the oil and gas industry is the Floating Production Storage
Offloading (FPSO) structure. FPSOs, as defined by Chakrabati, are ship shaped floaters [floating
structures] with provisions for storing and offloading of oil simultaneously (2005). FPSOs can be
considered to be in a quasi-static condition for two reasons. Firstly, FPSOs are able to weathervane
about its turret mooring, thus avoiding any large motions applied to the hull girder. Secondly, most
FPSOs have a detachable turret mooring, which allows the FPSO to detach and move to shelter if
unfavourable conditions present.
As the FPSO will be in a quasi-static condition, the ultimate strength of the hull girder becomes a
possible mode of failure. The ultimate strength of a floating structure is the maximum moment its hull
girder can sustain before collapse. This relationship can be plotted as a moment-curvature curve,
which shows the moment sustained by a hull for a corresponding curvature. A typical momentcurvature curve is shown in Figure 1.
The first attempts to determine the ultimate strength of a hull were investigated by Caldwell
(1965). In his paper, Caldwell developed the first method of estimating the ultimate longitudinal
strength of a vessel, firstly by considering yielding only for a cross section, then considering buckling
by introducing a buckling factor.
Caldwells work is of particular note as it is the seminal work for ultimate longitudinal strength of
vessels, and the method he proposed was simple and easy to implement. Caldwell made three
assumptions to reduce the complexity of the problem, the first being that plates that buckle will do so
uniformly, not non-uniformly as would be expected from a buckling plate failure. The second
assumption was that Caldwell used a ratio of the ultimate buckling strength to the ultimate yield
strength of the panel since at the time there were no methods to approximate the ultimate moment of
each panel. The third assumption is that there are only 4 panels present, one each for the upper and
lower decks, and two for the sides. These assumptions reduce the complexity of the problem
substantially.
Caldwells method is flawed however, and the issues stem from the assumptions made. Again,
Caldwell had no way of calculating the buckling load of plates, hence these simplifications made an
analysis possible, but methods have since been developed to address this. Also, reducing the problem
to four large panels may have reduced the complexity substantially, but does not represent the
behaviour of individual panels.
Smith (1977) developed a method to determine the moment-curvature relationship for a hull. In his
method, curvature is incrementally applied to the hull girder about the instantaneous neutral axis, and
by utilising the load-shortening curve for each panel, incremental stresses for each panel are found.
After this, the individual panel stresses are integrated over the cross section to find the incremental
bending moments applied.
Pre collapse
Elastic response
Post collapse
Residual strength
Methodology
Smiths method was implemented in MATLAB following the method Chen (2008) presents.
Chens implementation was selected as it presents the most comprehensive analysis of Smiths
method out of literature reviewed. The program itself can detect the curvature applied to cause a panel
to fail and presents a list of all panels with corresponding failure curvatures upon completion. The
program also tracks and exports the shift in the neutral axis as well as the moment-curvature
relationship.
The program is broken into a number of subroutines. The primary program, SmithsMethod.m,
controls assignments of curvature and zero matrices neutral axis and moment, reads the database with
panel information, initiates the neutral axis solver for hogging and sagging independently then finally
Panel Imperfections
Ship panelling in real-life conditions is not geometrically perfect, nor does it consist of perfect,
homogeneous material. Instead, panelling is susceptible to imperfections, including initial deflections,
residual stresses induced from welding, heat afflicted zones, corrosion, fatigue stresses and additional
loads beyond axial tension and compression. Each imperfection is considered in a different way
depending on the load-shortening curve model used and on the hull being analysed. Gordos model
considers residual stresses and initial deflections in his model by applying it as a reduction factor,
which reduces the ultimate load behaviour of the panel. Corrosion is taken as an effective reduction of
plate thickness, Rutherford and Caldwell (1990) consider a reduction of 0.1mm per year of service life
as an adequate allowance for steel vessels. Rutherford and Caldwell also considered corrosion to
affect both sides of a stiffener flange and hence applied the corrosion rate to both sides. Other effects
are not considered by Gordo and Guedes Soares, but can be assessed by finite element methods.
2.2
Hard Corners
Hard corners are located at the junction between major plates, Hu stating that it is thought
that the hard-corner element can fully withstand the load, that is to say, its load-end shortening curve
is the same as the stress strain curve of the elastic-perfectly plastic material. (2001). Hu later notes
that the plates comprising the corner are susceptible to buckling, and accounted for this with an
analytical equation. The amount of plate material that constitutes a hard corner varies with author, Hu
suggesting that a hard corner is simply half stiffener spacing from the corner, however Benson (2013)
considers the panel up to 30 plate thicknesses away from the junction to be a hard corner with an
elastic-perfectly plastic behaviour in compression. Bensons interpretation of what constitutes a hard
corner as well as his schema for element subdivision is shown in Figure 7.
Validation Results
For all programs a validation case was selected to confirm the program correctly predicts failure
and behaves as expected. The validation cases were selected by reviewing and selecting from case
studies that other authors have used, providing a comparable base line to compare results.
3.1
To show the validity of his methods, Hughes performed a number of comparisons with
experimental data. For his ultimate strain calculations, Hughes applied his method to a number of
experiments conducted on stiffened panels by Murray and Michelutti (1977). Murray and Micheluttis
experiments were conducted on panelling intended for box girder bridges, however the panelling
closely resembles panelling used for vessel hulls and therefore is comparable. The panels themselves
had the same nominal dimensions and stiffener shape, shown in Table 1. Hughes found that his
results correlated with Murray and Micheluttis results, with error of up to 16%.
Panel yield
stress
(MPa)
3.15E+08
3.96E+08
3.63E+08
3.67E+08
3.32E+08
3.70E+08
3.78E+08
3.68E+08
3.18E+08
3.24E+08
3.72E+08
3.70E+08
3.83E+08
3.78E+08
3.24E+08
Stiffener
yield stress
(MPa)
2.70E+08
3.79E+08
3.78E+08
3.70E+08
3.96E+08
3.81E+08
3.89E+08
3.83E+08
3.97E+08
3.85E+08
3.61E+08
3.38E+08
3.54E+08
3.71E+08
3.69E+08
Experimental
0
0
0.17
0.33
0.53
0.63
0.66
0.41
0.51
0.5
0.42
0.38
0.37
0.25
0
Hughes
Ultimate stress to
Yield stress ratio
0
0
0.15
0.31
0.51
0.64
0.59
0.39
0.51
0.46
0.38
0.33
0.31
0.24
0
Average error:
Error
0%
0%
12%
6%
4%
2%
11%
5%
0%
8%
10%
13%
16%
4%
0%
7%
Presented
Ultimate stress to
Yield stress ratio
0.00
0.05
0.18
0.39
0.56
0.66
0.65
0.44
0.51
0.44
0.40
0.38
0.37
0.35
0.20
Average error:
Error
0%
5%
3%
19%
6%
5%
1%
7%
1%
12%
5%
1%
1%
42%
20%
9%
Table 2 shows results from Hughes method to calculate panel ultimate strain. Each panels
ultimate buckling stress to yield stress ratio is compared to Hughes results and to experimental data.
The results for the presented method lie within an expected range of error for numerical results.
Notable outliers include panels Z, A5 and A1, with A1 showing 42% error. The variation in Panels
A1 is attributed to the program identifying the failure mode as Mode II, a buckling plate failure,
instead of the expected result of Mode III, the simultaneous failure of plate buckling and stiffener
yielding. Mode III is a rare case of failure, and it only occurs when the moment applied to the plate
approaches the plastic moment.
3.2
Normalised stress
0.7
0.6
0.5
0.4
Gordo results
0.3
Presented results
0.2
0.1
0
1
2
Normalised strain
3.3
787.4 mm
243.84 mm
4.88 mm
50.8 mm
4.8 mm
15.9 mm
4.8 mm
208.5 GPa
297.3 MPa
191.5 GPa
276.2 MPa
Gordo and Guedes Soares performed an analysis on a box girder which was tested to destruction
by Dowling (1973) . The box girder was divided into 18 panels for analysis; 8 identical panels in the
tension and compression flange, 4 panels with two additional plate sections for the web and 4 hard
corners. All panels were assessed with Gordos model, with the exception of the hard corners which
were treated as elastic-perfectly plastic elements. A cross section of Dowlings box girder #2 is shown
in Figure 9 below. Dimensions for the primary panel in the compressive and tensile flanges can be
found in Table 3 above, while dimensions for other panels can be found in relevant literature.
Figure 9 Cross section of Dowlings box girder (Gordo and Guedes Soares, 1996)
2.00E+09
1.80E+09
1.60E+09
Moment (Nmm)
1.40E+09
1.20E+09
Presented results
Gordo, 1996
1.00E+09
Dowling, 1973
8.00E+08
6.00E+08
Presented Hughes
estimate
4.00E+08
Hughes estimate
2.00E+08
0.00E+00
0.00E+00 1.00E-06 2.00E-06 3.00E-06 4.00E-06 5.00E-06
Curvature (1/mm)
Case study
With the lack of an FPSO cross section, VLCC Energy concentration presents an opportunity to
analyse a vessel with similar characteristics. Energy concentration was a crude carrier which broke its
back in Rotterdam in 1980 and has been analysed in detail since by multiple authors, such as
Rutherford and Caldwell. Rutherford and Caldwell performed a direct comparison between Smiths
method and the failure strength of the VLCC Energy Concentration determined by analysis of the still
water bending moment. Rutherford discussed the vessel and the circumstances that lead to its failure.
He also discussed Smiths method for ultimate strength analysis in more detail and made allowances
for other loads which would have contributed to the failure bending moment.
Figure 11 shows a comparison between VLCC Energy Concentration and an FPSO cross section.
While the FSPO is double bottomed, the VLCC provides a reasonable comparison. Both hulls exhibit
similar shapes, with comparable block coefficients (0.731 for the VLCC vs. 0.816 for the FPSO
shown in Figure 11). Both cross sections exhibit similarities, even while considering the presence of
2.50E+13
2.00E+13
1.50E+13
1.00E+13
5.00E+12
Gordo results
0.00E+00
-6.00E-07 -4.00E-07 -2.00E-07 0.00E+00 2.00E-07 4.00E-07 6.00E-07 8.00E-07
No corrosion allowance
Corrosion allowance
-5.00E+12
-1.00E+13
-1.50E+13
-2.00E+13
-2.50E+13
Curvature (1/mm)
Conclusion
Ultimate strength in FPSOs is an issue that needs to be considered for their entire operational
lives. Two methods for determining hull ultimate strength were selected, a variant of Smiths method
implemented by Chen and a method derived from Smiths developed by Hughes. These methods were
selected because of the advantages they present over finite element methods lower computational
time, greatly simplified set-up and lower software costs. These methods were programmed in
MATLAB with methods verified against past experimental data and previous work. Verification cases
selected were panel ultimate strength experiments conducted by Murray and Michelutti, a loadshortening curve generated by Gordo, ultimate strength of a box girder tested to destruction by
Dowling and ultimate strength of VLCC Energy Concentration.
Hughes method for predicting panel ultimate strain shows that results generated were within the
same margins of error as those found by Hughes with the exception of one result, a rare case of
extreme failure. Hughes method for calculating ultimate strength was found to have an error of 24%,
a large margin of error. Comparisons made with Gordos load-shortening curve suggest that Gordos
method is implemented correctly without residual stresses or other imperfections for beam-column
failure. The importance of residual stresses and imperfections is highlighted with a comparison to the
moment-curvature curve found experimentally by Dowling, results presented show an 11% increase
Future recommendations
The following issues should be investigated to improve the accuracy of methods proposed above.
A procedure for generating reliable load-shortening curves with finite element software which can
account for imperfections and other non-axial loads would increase the accuracy of results, with a cost
to computational speed. Hard corners should be reviewed to determine their behaviour when under
compressive axial loadings, along with a method to determine to what extent an element in a cross
section is a hard corner. The use of elastic-perfectly plastic stress strain curves to represent elements
in tension should be reviewed, as this is at best a conservative estimate.
References